Parametric Analysis on Crop/Weed Classification via Optimized Convolutional Neural Network

• Published in: 2021 Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV) pp. 1477 - 1484
• Main Authors: Panda, Bikramaditya, Mishra, Manoj Kumar, Mishra, Bhabani Shankar Prasad, Tiwari, Abhinandan Kumar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 04-02-2021
• Subjects: agriculture; Classification algorithms; Convolutional Neural Network; Convolutional neural networks; Feature extraction; Gray-Level Run- Length Matrix; Hybridized Whale and Sea Lion Algorithm; Support vector machines; Training; Tuning; Weed Control; Whales
• DOI: 10.1109/ICICV50876.2021.9388462

A new crop/weed classification model is established in this work that includes three main phases like (1) Preprocessing (2) Feature extraction (3) and Classification Initially, the input image is pre-processed via contrast enhancement process. Subsequently, feature extraction is performed, where "Gray Level Co-occurrence Matrix (GLCM) as well as Gray-Level Run-Length Matrix (GRLM)" based features are extracted. These extracted features together with RGB image (totally 5 channels) are classified via "optimized Convolutional Neural Network (CNN) For enhancing the classification, the weight and activation function of CNN are chosen optimally via Hybridized Whale and Sea Lion Algorithm (HW-SLA) model. Eventually, algorithmic analysis is carried out on proposed HW-SLA algorithm by varying its parameters.